Re: generic plans and "initial" pruning

Hi Rui,

> I've combined Manu's correction and the test in the attached patch,
> which applies on top of v13.

Thanks for folding in the test -- it fails on v13 without the ownership
correction and passes with it, the same as I saw.

cfbot is red only because v13 itself needs a rebase: 1b5dd3a24a4 added an
early return in InitExecPartitionPruneContexts() that conflicts with
v13-0002. The resolution is mechanical -- the lines v13-0002 adds and
removes are unchanged, only the surrounding context moved, and the
patch-ids of v13-0001/0003/0004 and your 0001 are identical before and
after.

On a v13 rebased that way on today's master (0d6103e537e1) with your 0001
on top: make check 239/239, isolation 134/134 and test_plan_advice all
pass. Without the ownership fix, make check fails one test with
"plancache reference ... is not owned by resource owner Portal", so the
fix is still doing its job on current master.

I'm happy to post the rebased series if that's useful; otherwise it's a
one-liner of context for whoever rebases v13.

Regards,
Manu